Open DanielHabenicht opened 1 month ago
Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @gulopesd @Haiying-MSFT @jairmyree @joshfree @KarishmaGhiya @KevinBlasko @kurtzeborn @nisha-bhatia @pvaneck @sarangan12 @scottaddie @srnagar @ToddKingMSFT.
@xiangyan99
I do not own `azure-monitor-query, please reassign.
Hey @DanielHabenicht. The azure-monitor-query SDK only outputs what the monitor metrics API returns, so it's likely a service-side issue. If you were to view the OpenAI metrics in the Azure Portal, does it show the same? I'm not sure what's up, but it seems like some of the dimensions aren't being populated/leveraged for the TokenTransaction
metric. Couldn't find much online regarding the issue (only really this question for something similar for Cosmos metrics).
You'd probably get better eyes on this by opening a question on the Microsoft Q&A website or opening a support request (can do this by going to the "Help + support" option in the left-hand menu of your OpenAI resource in the Azure Portal).
In the Azure Portal the TokenTransaction
Metric is not available :/
I will ask the support directly and update this issue accordinly.
Hi @DanielHabenicht. Thank you for opening this issue and giving us the opportunity to assist. To help our team better understand your issue and the details of your scenario please provide a response to the question asked above or the information requested above. This will help us more accurately address your issue.
Based on these docs TokenTransaction
corresponds to Processed Inference Tokens
. After that, can apply splitting on ModelName and ModelDeploymentName
Based on these docs
TokenTransaction
corresponds toProcessed Inference Tokens
. After that, can apply splitting on ModelName and ModelDeploymentName
Yes, somehow I must have overlooked that. It shows up in the Azure Portal as well:
azure-monitor-query
1.4.0
mcr.microsoft.com/devcontainers/python:1-3.11-bookworm
) on Windows 10Describe the bug The
modeldeploymentname
is empty for certain deployed models. NamelyDALLE3.Standard.Low
andWhisper
To Reproduce Query the metrics of the specified models:
Expected behavior It should contain the model name as specified during the deployment, instead of empty.
Additional context I can provide the exact resource id via PN.